Chongqing is located in mountainous area where the vegetation is rich. It is greatly affected by the fog all year round. As a result, the noise of remote sensing image is serious and the affected area is large which have a great influence on the accuracy of related application.NDVI is one of the most important vegetation index which has an important role in the study of global climate change, vegetation change etc. In this paper, the MODIS/NDVI data are used, the time resolution of the data is monthly and the spatial resolution is lkm. The NDVI data of Chongqing city from 2010 to 2014 are reconstructed by four smoothing techniques: WS, S-G, A-G and D-L. The results are evaluated by visual comparison, pixel contrast of different land use types,correlation coefficient(R),the root mean square error (RMSE), Akaike Information Criteria (AIC) and Bayesianlnforma- tion Criteria (BIC). In the visual contrast analysis, the results indicate that the noise of S-G technique and WS technique is lower than the others. In the pixel contrast of different land use types, the curve of S-G technique is the smoothest.In the fidelity analysis, R and RMSE are used to evaluate of the original data and data reconstructed by four technique, the result of A-G technique and WS technique is better. It also indicates that the A-G technique have the widest distribution, the value of R larger than 0.8 accounts for 89.41% of the total area and the value of RMSE smaller than 0.05 accounts for 66.40% of the total area. The WS technique is better than the other two techniques, accounting for 72.76% and 59.37% of the total area respectively. In the model effect analysis, the AIC and BIC evaluation results of the A-G technique are the best. The evaluation results of WS techniqueis good in Western Chongqing where the BIC evaluation results of other threetechniques are relatively poor, the result of S-G method is the worst.